Role of the COOH-terminal Domains of Meprin A in Folding, Secretion, and Activity of the Metalloendopeptidase*

Secreted forms of the a subunit of recombinant mouse meprin A include an NH 2 -terminal prosequence, a cata- lytic domain, and three COOH-terminal domains desig-nated as MAM (meprin, A-5 protein, receptor protein- tyrosine phosphatase m ), MATH (meprin and TRAF homology), and AM (after MATH). In this study, the importance of these COOH-terminal domains for biosynthesis of secreted, activable forms of the protease was investigated. Transcripts of the meprin subunit truncated after the protease ( a (1–275)), MAM ( a (1–452)), and MATH ( a (1–528)) domains or with individual domains deleted ( D MAM, D MATH, and D AM), were transfected into human embryonic kidney 293 cells. The wild-type subunit, D MATH, D AM, a (1–452), and a (1–528) were secreted into the media, although the D AM mutant was secreted at very low levels. The D MATH and a (1–452) mutants were not activable by limited proteolysis. The a (1–528) mutant was as active as wild-type meprin a against a bradykinin substrate, but had no activity against azocasein, and it, as all other mutants, was more vulnerable to extensive degradation by proteases than the wild-type protein. Pulse-chase experiments revealed that the D MAM and a (1–275) mutants were rapidly degraded within cells. Treatment with lactacystin, a specific inhibitor of the proteasome, significantly II-S) added at a 2-fold excess of trypsin. incubated at 25 °C for 15 and 4 °C proteolytic susceptibility studies, trypsin was added the secreted proteins at or ng/ m and incubated 37 for 30 Endoproteinase used to acti- vate meprin subunits a of ng/ m 37 °C, m M 3,4-dichloroisocoumarin.

units that contain astacin-like catalytic domains and disulfidebridged dimers (6 -10). They are capable of degrading proteins such as collagen and gelatin, hormones such as parathyroid hormone, luteinizing hormone-releasing hormone, and melanocyte-stimulating hormone, and small peptides such as bradykinin, angiotensins, and gastrin (11,12). The meprins may, therefore, be involved in activation or inactivation of important extracellular proteins and peptides, and are highly regulated at transcriptional and posttranscriptional levels themselves. They are tissue-specific proteases that are implicated in developmental processes, as well as in normal and pathological processes in adult tissues (1,13).
Studies of recombinant forms of meprins expressed in mammalian cells have yielded important information about critical features of the enzyme (8 -10, 14, 15). For example, it has been demonstrated that removal of a 56-amino acid domain in the ␣ subunit (the I domain; see Fig. 1) results in retention of the subunit in the endoplasmic reticulum (ER), 1 and that mutation of Cys 320 in the MAM domain to Ala results in a monomeric secreted, activable at form of the protease that is unstable to heat and proteolytic degradation. Attempts to express enzymatically active or activable forms of the recombinant protease domain of meprin or the protease domain with signal and/or prosequences (in the absence of COOH-terminal domains) have been unsuccessful, in that these proteins are either degraded in the cells or accumulate in inclusion bodies depending on the cell type. 2 These results implied that the noncatalytic, COOH-terminal domains of meprins are important for the secretion and correct folding of the enzyme, although this has not been specifically investigated previously.
The deduced amino acid sequences of meprin ␣ and ␤ subunits from mouse, rat, and human cDNA have been determined (Refs. 3, 4, 6, and 16; see Fig. 1). The subunits encode an amino-terminal signal sequence (S), a prosequence (Pro), a protease (astacin-like) domain, a MAM (meprin, A-5 protein, receptor protein-tyrosine phosphatase ) domain, a MATH (meprin and TRAF homology) domain, an AM (after MATH) domain, an epidermal growth factor-like domain, a putative transmembrane-spanning (T) domain, and a cytoplasmic (C) domain (15). The domain structure of the meprin ␣ and ␤ subunits are similar except that meprin ␣ encodes a 56-amino acid sequence (the I or inserted domain) located between the AM and the epidermal growth factor-like domains that is absent in the ␤ subunit (9). The I domain enables COOH-terminal proteolytic processing of the meprin ␣ subunit at residue Arg 615 (near the end of the AM domain) in the ER, so that this subunit is secreted if not associated with a ␤ subunit at the cell surface (9,15). The MAM and MATH domains occur in several other proteins, and have been implicated as "interaction" or "adhesion" domains (17)(18)(19)(20). The MAM domain, which is present in cell surface proteins, consists of about 170 amino acids, contains four or five cysteine residues, and has been shown to play a role in homodimerization of protein-tyrosine phosphatase (17,18). The MATH domain is found in cytosolic proteins such as TRAF (tumor necrosis factor receptor-associated factor) and has been shown to be essential for homodimerization and heterologous interactions with receptor proteins (19,20). Functions of the AM domain are unknown, and no homologous sequences have been identified for this domain in proteins other than meprin.
In the present study, we investigated the role of the MAM, MATH, and AM domains in secretion, folding, and activity of the meprin ␣ subunit by mutational (see Fig. 1) and transfection analyses.

EXPERIMENTAL PROCEDURES
Reagents and Materials-[ 35 S]Methionine/cysteine was purchased from ICN Pharmaceuticals, Inc. Lactacystin and Pansorbin were from Calbiochem. The pcDNA I/Amp plasmid was from Invitrogen. Dulbecco's modified Eagle's medium, methionine-free Dulbecco's modified Eagle's medium, and Opti-MEM were from Life Technologies, Inc. The transformer site-directed mutagenesis kit was from CLONTECH. Protein A-Sepharose was from Sigma. Endoglycosidase H (Endo-H) and endoglycosidase F/N-glycosidase F (Endo-F) were from Boehringer Mannheim.
Plasmid Construction and Mutagenesis-The pcDNA I/Amp plasmid expressing full-length wild-type mouse meprin subunit cDNA was described previously (9,10). The COOH-terminal truncation and deletion mutants ( Fig. 1) were generated by the polymerase chain reaction (PCR) by methods described previously (15) with some modifications. The truncation mutants were constructed by PCR using mutagenic antisense primers. The ␣(1-275) transcript was constructed by changing the codon of the wild-type transcript at Thr 276 to a stop codon. For the ␣(1-452) transcript, the codon for Ile 453 was changed to a stop codon. For the ␣(1-528) transcript, Ala 529 was changed to a stop codon. The deletion mutants were constructed by the methods of Higuchi using primers containing 5Ј add on sequence overlapping sequences (21). The ⌬AM mutant was constructed by fusing the NH 2 -terminal fragment of Met 1 -Glu 528 with the COOH-terminal fragment of Arg 628 -Glu 760 using primers in the PCR amplification such as 5Ј-CATAATGACCATCCTG-GACCAGGAAAGAGGCCTCCTTCTGCAAG3Ј and 5Ј-CTTGCAGAAG-GAGGCCTCTTTCCTGGTCCAGGATGGTCATTATG-3Ј. The ⌬MATH mutant was constructed by fusing Met 1 -Ala 445 with Arg 529 -Glu 760 using the PCR amplification primers 5Ј-CTGACAGAAACCCCCTGCC-CTGCA-GCTGATACCAGGAACAGGATGTCC-3Ј and 5Ј-GGACATCC-TGTTCCTGGTATCAGC-TGCAGGGCAGGGGGTTTTCTGTCAG-3Ј.
The ⌬MAM mutant was constructed with Met 1 -His 275 and Gly 446 -Glu 760 using the PCR amplification primers 5Ј-TACAACTGCACCGC-AACACATGGGGTTTGGACCACTCGGAAATA-3Ј and 5Ј-TATTCCGG-ATGGTCCAAACCCCATGTGTTGCGGTGCAGTTG-3Ј. These NH 2 -terminal and the COOH-terminal fragments were separately constructed, and then both fragments were extended with Pfu DNA polymerase (Stratagene). All PCR reactions were constructed with Pfu DNA polymerase to minimize base misincorporations. All constructs were verified as correct by DNA sequence analyses.
Tissue Culture and Transfection-Human embryonic kidney 293 cells (ATCC 1573 CRL) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 50 units/ml penicillin, and 50 g/ml streptomycin (complete Dulbecco's modified Eagle's medium) in a 37°C incubator with 5% CO 2 . The recombinant meprin ␣ wild-type or mutants were expressed in human embryonic kidney 293 cells after transfection by the calcium phosphate precipitation method using 10 g of expression plasmid and 1 g of helper plasmid pVA1/ 100-mm tissue culture plate (9). Cells were grown to approximately 90% confluence by overnight incubation in complete Dulbecco's modified Eagle's medium. Then the medium was replaced with serum-free Opti-MEM (5 ml/plate) and returned to the incubator for an additional 48 h.
Preparation of Media and Cell Lysates-The tissue culture media were collected after 48 h of transfection and were subjected to centrifugation at 16,000 ϫ g for 20 min. The supernatant fractions were concentrated 10-fold to 500 l/plate using Centriprep-30 and Microcon-30 concentrators (Amicon, Inc.). The cells were washed twice with phosphate-buffered saline (PBS), removed from the plates with a rubber scraper, and subjected to centrifugation at 300 ϫ g for 5 min. The sedimented cells were suspended in PBS containing 0.1% Triton X-100, sonicated for 1 min at 4°C, and subjected to centrifugation at 100,000 ϫ g for 1 h; the supernatant fraction is referred to as the cell lysate.
Immunoprecipitation of Labeled Meprins-The cells and media were mixed with 40 l of Pansorbin for 1 h at 4°C to prevent nonspecific binding to IgG-protein A complexes, and then centrifuged at 6,500 ϫ g for 30 min. The supernatant fractions were incubated with 15 l of anti-mouse meprin ␣ IgG at 37°C for 10 min, and then stored at 4°C for 16 h. The immunocomplexes were mixed with 40 l of Protein A-Sepharose beads (50% gel suspension) for 3 h at 4°C with gentle agitation. The beads were washed three times with 0.1% SDS, 0.1% Triton X-100, 200 mM EDTA, 10 mM Tris-HCl (pH 7.5), washed another three times with the same buffer containing 1 M NaCl and 0.1% sodium lauryl sarcosinate, and then washed twice with 5 mM Tris-HCl (pH 7.0). The sedimented beads were boiled for 5 min at 100°C with 50 ml of 0.1% SDS, 0.5 mM EDTA, 5% sucrose, 5 mM Tris-HCl (pH 8.0) with 2-mercaptoethanol.
SDS-PAGE and Immunoblotting-The supernatant fractions were subjected to electrophoresis using 7.5% SDS-polyacrylamide gels (22). Immunoblotting was performed as described previously (9). The proteins were probed with one of two different antibodies, which had been produced by injection of meprin antigen into rabbits. In one instance, the purified mouse meprin had been deglycosylated prior to injection (referred to as anti-deglycosylated antibody); the second antibody was produced in response to native meprin antigen (anti-␣ antibody). Meprin subunits were detected using the enhanced chemiluminescence method (Pierce).
Endoglycosidase Digestion-The immunoprecipitated or secreted proteins were denatured by boiling for 5 min at 100°C with 5 mM Tris-HCl (pH 8.0) containing 0.2% SDS. The samples were adjusted to a final concentration of 50 mM sodium acetate buffer (pH 6.0), containing 0.75% Triton X-100 and 100 g/ml protease inhibitors such as antipain, chymostatin, leupeptin, pepstatin, and phenylmethylsulfonyl fluoride. The mixtures were added to 10 milliunits of Endo-H or 1.0 unit of Endo-F, and incubated at 37°C for 18 h. Reactions were stopped by boiling the samples in SDS-PAGE sample buffer.
Trypsin and Arg-C Treatment-Trypsin was added to the secreted proteins at the final concentration of 10 ng/l in 20 mM Tris-HCl, pH 7.5. After incubation at 25°C for 30 min, soybean trypsin inhibitor (Sigma, type II-S) was added at a 2-fold excess of trypsin. Samples were incubated at 25°C for 15 min and then kept at 4°C until analysis. For proteolytic susceptibility studies, trypsin was added to the secreted proteins at 10, 20, 30, or 40 ng/l, and then incubated at 37°C for 30 min. Endoproteinase Arg-C (Boehringer Mannheim) was used to activate meprin subunits at a concentration of 40 ng/l. After 30 min at 37°C, the reaction was stopped with 0.3 mM 3,4-dichloroisocoumarin.

RESULTS
Wild-type Meprin ␣ and Mutants ␣(1-528), ␣(1-452), and ⌬MATH Were Secreted into the Media-When the wild-type mouse meprin ␣ subunit cDNA was transfected into human embryonic kidney 293 cells, the protein was secreted into the medium with a molecular mass of approximately 95 kDa ( Fig.  2A, lane 2). A small amount of the expressed meprin ␣ subunit was found associated with cell lysates (Fig. 2A, lane 1); this component had a molecular mass of approximately 90 kDa. Truncated mutants ␣(1-528), ␣(1-452), and the ⌬MATH deletion mutant were also detected as immunoreactive, secreted proteins with molecular masses of approximately 90, 75, and 92 kDa, respectively. By contrast, mutants ␣(1-275), ⌬MAM, and ⌬AM were undetectable by immunoblot analysis of the cell lysates and media ( Fig. 2A, lanes 7-10, 13, and 14). Repeated analyses revealed that the expression levels of all the mutants in the media were lower than the wild-type, but deletion of the MAM domain (⌬MAM), truncation just before the MAM do-main (␣(1-275)), and deletion of the AM domain (⌬AM) yielded the lowest levels of secreted protein (Fig. 2B).
When the secreted proteins were subjected to SDS-PAGE in the absence of ␤-mercaptoethanol, all migrated as dimers (data not shown). Thus, the intersubunit S-S bridging was maintained in secreted mutants as in the wild-type.
The Secreted Proteins Contain Complex-type Oligosaccharide Chains-To determine whether the secreted mutants contained complex oligosaccharides as the wild-type (9), they were treated with Endo-H and Endo-F (Fig. 3). The ␣(1-528), ⌬MATH, and ␣(1-452) proteins were all resistant to Endo-H, and sensitive to Endo-F, indicating complex glycosylation had occurred. Endo-F treatment decreased the molecular mass of the wild-type from about 95 to 75 kDa (Fig. 3, lanes 1 and 3); the ␣(1-528) mutant from approximately 90 to 70 kDa (Fig. 3,  lanes 4 and 6); the ⌬MATH mutant was from about 92 to 72 kDa (Fig. 3, lanes 7 and 9); and the ␣(1-452) mutant was shifted from approximately 75 to 65 kDa (Fig. 3, lanes 10 and  12). The deglycosylated proteins all migrated slower on gels than their predicted molecular sizes (by about 10 kDa), but as predicted relative to each other. Endo-F treatment decreased the molecular masses of the wild-type, ␣(1-528), and ⌬MATH mutants approximately 20 kDa, as assessed by mobility after SDS-PAGE. The ␣(1-452) mutant lost approximately 10 kDa upon treatment with Endo-F. The decreased mass lost from the latter mutant relative to others may be due to the absence of the MATH and AM domains, both of which contain potential glycosylation sites.
The MATH Domain Is Required for Biosynthesis of an Activable Form of Meprin-Meprin ␣ subunits are secreted as inactive proenzymes and must be treated with trypsin-like proteinases for the conversion of the zymogen to the active   FIG. 2. Expression of wild-type, truncation, and deletion mutants of meprin ␣ in human embryonic kidney 293 cells. A, human embryonic kidney 293 cells were transfected with cDNA for wild-type meprin ␣ or mutants. The cell lysates (designated C) and culture media (designated M) were prepared as described under "Experimental Procedures." Both samples were subjected to SDS-PAGE (7.5% gels with ␤-mercaptoethanol), followed by immunoblotting using anti-mouse meprin ␣ polyclonal antibodies. B, the amount of meprin ␣ protein per microliter in both the cell lysates and the media was determined by densitometric scans of immunoblots using laser densitometry as described previously (12). The data are shown as the mean Ϯ S.E. of values determined in five independent experiments. Wt, wild-type. mature enzyme (11)(12)(13)(14)(15). To test whether secreted mutants ␣(1-528), ␣(1-452), and ⌬MATH were correctly folded proteins and activable enzymes, the mutants were treated with trypsin and Arg-C, under conditions used for limited proteolysis of the wild-type enzyme, and catalytic activities were measured against a bradykinin analog (BKϩ) and azocasein. After incubation with 40 ng/l endoprotease Arg-C at 37°C for 30 min, the wild-type and the secreted mutants ␣(1-528), ␣(1-452) and ⌬MATH migrated faster on polyacrylamide gels than nontreated proteins, indicating limited proteolysis (Fig. 4A). Prior to treatment with Arg-C neither wild-type nor mutants had activity. After the limited proteolysis, the wild-type and ␣(1-528) mutant proteins had comparable activities against the bradykinin substrate. However, the mutants lacking the MATH domain (⌬MATH and the ␣(1-452) mutant) had no detectable activity (Table I). Activity of all mutants against azocasein was markedly decreased. The ␣(1-528) mutant had a 10-fold decreased activity against the protein substrate compared with the wild-type, and no activity of the mutants lacking MATH domain could be detected with azocasein. These results indicate that the mutants were synthesized as proforms, that the MATH domain is necessary for synthesis of an enzymatically activable protease (against peptide and protein substrates), and that the AM domain is necessary for generation of activity against protein substrates but not against small peptides such as bradykinin.
All Mutants Are More Susceptible to Degradation by Trypsin than the Wild-type-Notable differences in the vulnerability to extensive degradation by trypsin were also observed between the wild-type and secreted mutant proteins. The wild-type and ␣(1-528) mutant were resistant to general proteolysis when treated with 10 ng/l trypsin, while limited proteolysis at this trypsin concentration shifted the proforms to the enzymatically active forms (Fig. 4B, lanes 1-4). However, under the same conditions, the ⌬MATH and ␣(1-452) mutants were extensively degraded (Fig. 4B, lanes 5-8). Differences in the susceptibility between the wild-type and the ␣(1-528) mutant were found when they were treated with a higher concentration of trypsin (Fig. 4C). The ␣(1-528) mutant was extensively degraded by 20 -40 ng/l of trypsin (Fig. 4C, lanes 7 and 8), whereas the wild-type was resistant to proteolysis (Fig. 4C,  lanes 3 and 4). These results indicate that the mutants lacking the MATH domain were not correctly folded to generate active enzymes, while the mutant lacking the AM domain (␣(1-528)) had a correctly folded protease domain for generation of peptidase activity, but lacked proper structure for activity against protein substrates and resistance to degradation.

Pulse-Chase Experiments Reveal That the ⌬AM Mutant Was Secreted at Low Levels, and That ⌬MAM and ␣(1-275) Were
Rapidly Degraded within Cells-To determine whether mutants undetectable by immunoblot analysis (see Fig. 2) were synthesized and rapidly degraded, pulse-chase experiments were performed (Fig. 5). The wild-type protein associated with cells was initially observed as a protein of approximately 100 kDa, and then converted to a 95-kDa form; by 2 h after the FIG. 3. Deglycosylation of the secreted wild-type and mutant meprins. The secreted meprin subunits in media were incubated at 37°C for 18 h with or without Endo-H (H) and Endo-F (F). The mixtures were subjected to SDS-PAGE, and analyzed by immunoblotting using anti-meprin ␣. Wt, wild-type.

FIG. 4. Limited and extensive proteolysis of the secreted wildtype and mutant meprins.
A, the secreted wild-type (Wt) and mutants were incubated at 37°C for 30 min with or without 40 ng/l protease Arg-C. The Arg-C was then inactivated with 0.3 mM 3,4dichloroisocoumarin. B, the secreted wild-type and mutants were incubated at 25°C for 30 min with or without 10 ng/l trypsin. The samples were then inactivated with a 2-fold excess of soybean trypsin inhibitor. C, the secreted wild-type and mutants were incubated at 25°C for 30 min with 10, 20, and 40 ng/l trypsin. The incubated samples were then inactivated with a 2-fold excess of soybean trypsin inhibitor. The mixtures were subjected to SDS-PAGE followed by immunoblotting.

TABLE I Specific activity of secreted wild-type and mutant meprins
Secreted wild-type and mutants were activated with endoproteinase Arg-C as described under "Experimental Procedures." Activity was determined using 24 M fluorogenic bradykinin analog (BKϩ) or 11 g/l azocasein as substrates. The amount of meprin protein was estimated from immunoblotting analyses using a purified mouse meprin A preparation as a standard as described previously (10). Activities are expressed as the mean Ϯ S.E. ND, not detectable. radiolabeled pulse, the protein was detectable in the medium (Fig. 5A). All the mutants that were undetectable by immunoblot analysis were detected in cells after the radiolabel pulsetime was increased from 10 to 30 min. The radiolabeled ⌬AM mutant was also detected in the medium; the cell-associated and secreted proteins were approximately 88 kDa (Fig. 5C). This indicates that the protein can fold and be transported without the AM domain but that the protein is inefficiently secreted without this domain. The ␣(1-275) mutant was observed as two polypeptide forms of approximately 45 and 42 kDa after the 30-min radiolabeling period (Fig. 5B). After 2 h, only a small amount of the protein was retained within the cell, and no protein was detected in the medium (Fig. 5B, lane 2). Similarly, the ⌬MAM mutant was initially synthesized with a molecular mass of about 85 kDa, and then rapidly disappeared within the 2-h chase period (Fig.  5D). Over a 10-h period of incubation with unlabeled amino acids, no radiolabeled protein was secreted in the medium (Fig.  5D, lanes 5-8). These results indicate that the mutants ␣(1-275) and ⌬MAM were degraded within the cell.
To determine whether the cell-associated mutant proteins were retained in a pre-Golgi compartment, deglycosylation experiments were performed. After a 2-h pulse with radiolabeled amino acids, the ⌬MAM and the ␣(1-275) were deglycosylated with Endo-H and Endo-F. The proteins of these mutants were sensitive to Endo-H and Endo-F, indicating that they were high mannose-type forms (Fig. 6A). Moreover, when cells were treated with brefeldin A, a compound that prevents the transport in the secretory pathway from the ER to the Golgi complex (23), the degradation of the mutants of the ⌬MAM and the ␣(1-275) mutants was not inhibited (data not shown). These results indicate that degradation of the mutants ⌬MAM and ␣(1-275) occurs in a pre-Golgi compartment. When the ⌬AM mutant was treated with Endo-H and Endo-F, the cell-associated protein was sensitive to both endoglycosidases. The 88-kDa protein was converted into two forms of approximately 70 and 68 kDa (Fig. 6B). However, the secreted ⌬AM protein was resistant to Endo-H, indicating that it had acquired complextype oligosaccharide chains.
The vealed that mutants of several membrane-bound and soluble proteins that enter the secretory pathway are degraded by the proteasome located in the cytosol (24,25). To examine whether the proteasome is involved in the degradation of the ⌬MAM and ␣(1-275) mutants, cells were incubated with lactacystin, a specific inhibitor of the proteasome (26). In the presence of lactacystin, the rates of degradation of the ⌬MAM and ␣(1-275) mutants were significantly decreased (Fig. 7, B and C). Treatment with lactacystin resulted in retention of the ⌬MAM mutant within the cells for at least 5 h and the ␣(1-275) mutant for 2 h (Fig. 7, B and C). Under the same conditions, treatment of cells with lactacystin had no notable effect on the biosynthesis or secretion of the wild-type protein (Fig. 7A). These results indicate that the mutants lacking the MAM domain are degraded at least in part by the proteasome. DISCUSSION The work herein clearly demonstrates that the MAM, MATH, and AM domains of meprin ␣ are essential for efficient transport of the protein to the cell surface and/or correct folding to generate an enzymatically active protease. A summary of the effects of truncations of the subunit and deletions of the domains on expression, activity, and protease susceptibility is shown in Table II. The only mutant in this series that was secreted (at levels detectable by immunoblotting) and enzymatically activable was ␣ (1-528), the truncated mutant missing COOH-terminal domains after the MATH domain. The fact that this mutant displayed full activity against bradykinin, but very little activity against azocasein, is reminiscent of truncated mutants of MMP-13 in which the COOH-terminal hemopexin domain is removed (27). The latter mutant can be activated as the wildtype for activity against peptide substrates, but has no activity against collagen. For the MMPs, the hemopexin domain is critical for interaction with the collagen substrate, and in an analogous fashion, it is possible that the AM domain serves this type of function in meprin for protein substrates. In previous studies, monomeric forms of meprin, produced by mutations of cysteine residues in the MAM domain that prevent intersubunit disulfide bridges, were found to display this characteristic of activity against peptides but not proteins, which led to the suggestion that the oligomeric structure of the enzyme was important for interaction of the enzymes with protein substrates (10). The data herein provide further evidence that meprin activity against proteins is determined by domain-domain interactions that are critical for structure and full function of the enzyme.
Previous studies have shown that the meprin ␣(1-570) mutant, truncated near the end of the AM domain, was secreted at levels comparable to the wild-type protein (10). The ␣(1-528) mutant in the present study was secreted at a lower level, approximately 20% of wild-type. Thus, residues 528 -570 of the AM domain appear to be important for the efficient secretion of the subunit. This segment of the AM domain contains at least one glycosylated residue (Asn 546 and/or Asn 554 ), determined from carbohydrate analyses of mouse meprin ␣ (3), and may be important for interaction with chaperones or other factors in the ER involved in transport from the ER to the Golgi apparatus. The loss of the oligosaccharide chain in mutants missing the AM domain may also increase the susceptibility of the protein to proteolytic degradation.
The ⌬AM mutant was secreted at an even lower level than ␣(1-528); ⌬AM was only detectable with radiolabeling techniques. The difference between the ⌬AM mutant and ␣(1-528), is that the former is synthesized with all domains COOHterminal to AM, and because of this contains a transmembrane-spanning segment. Previous studies have established that the COOH terminus of the secreted mouse kidney meprin ␣ subunit is at Arg 615 in the AM domain, and that the I domain is essential for the COOH-terminal proteolytic processing of the wild-type protein to occur in the ER (9,15). Deletion of the AM domain may also impede the efficient COOH-terminal proteolysis of the subunit, and thereby retard its movement through the secretory pathway. Taken together, all the data indicate that the AM domain plays a role in secretion due to its importance in COOH-terminal proteolytic processing and glycosylation of the subunit.
Mutants lacking the MATH domain are folded sufficiently for secretion (at approximately 40 -70% of the wild-type level), but not correctly to generate an active protease from the proform. In addition, both the ⌬MATH and MATH truncation mutant ␣(1-452) were very vulnerable to extensive degradation by trypsin, which is another indication of incorrect folding. Meprins are dependent on external proteases for removal of the prosequence; they do not have a cysteine switch mechanism for activation (28), as the matrixins, and cannot be auto-or selfactivated (14). The MATH domain likely plays a role in domaindomain interactions that allow limited proteolysis, for removal of the prosequence, and also prevent extensive proteolysis of the subunit. It has been suggested that prosequence removal in meprins allows the formation of hydrogen bonds involving the two NH 2 -terminal residues that are critical for enzyme structure (14). The MATH domain may be important for correct dimerization or oligomerization of the enzyme, as it plays this role in other proteins. There is approximately 30% identity in the MATH domain between TRAFs and meprins (19). Deletion of the MATH domain of CRAF (CD40 receptor-associated factor 1), which is a relative of TRAF, revealed that the MATH domain in the cytosolic proteins is essential for homodimerization and heterologous interactions with receptor proteins in acute phase responses, lymphocyte activation, and nerve cell growth (20). It is notable that the MATH domain is conserved between TRAFs and meprins, since TRAFs are intracellular binding proteins in the cytosol whereas meprins are extracellular metalloproteinases. However, there are no cysteine residues in the MATH domain, therefore folding does not depend on the reducing/oxidizing environment.
The MAM domain in meprin ␣ appears to be particularly important for correct folding of the protein during biosynthesis, as without this domain the protein is degraded via a proteasomal route. These data imply that meprins are members of a growing group of membrane and secreted proteins in which chaperones bind to misfolded proteins during biosynthesis, and facilitate retrograde transport to the cytosol for degradation by the proteasome (24,25). Examples of this group include unassembled subunits of T cell receptor (29), a mutant of the cystic fibrosis transmembrane conductance regulator (30), 3-hydroxy-3-ethylglutaryl-CoA reductase (31), apolipoprotein B100 (32), and ␣ 1 -antitrypsin Z (33). Wiertz et al. (34) demonstrated that misfolded proteins are exported across the ER membrane to the cytosol for degradation through a proteinaceous channel, the Sec61p translocon. In meprin mutants containing the MAM domain, chaperones may interact with folding intermediates and facilitate formation of correct conformations that lead to COOH-terminal proteolytic processing and transit through the ER. When the MAM domain is not present, chaperones may bind to misfolded protein and enable retrograde transport of the protein from the ER to the cytoplasm.
Studies with a recombinant form of human meprin ␣, ␣(1-325), which was truncated after the first 62 amino acids residues of the MAM domain, so that approximately one third of the MAM domain was present in the protein, indicated that this mutant was secreted efficiently from Madin-Darby canine kidney cells (35). Thus, it is possible that only the first third of the MAM domain is critical for secretion. The first third of the MAM domain contains three cysteine residues, and studies with mouse and rat meprin ␣ transcripts have indicated that one of those residues (human Cys 308 , eqivalent to Cys 320 in the mouse and Cys 309 in the rat) is responsible for intersubunit S-S bridging. The folding of the MAM domain during biosynthesis and covalent dimerization of meprin molecules could impact on the interaction of the protein with chaperones and other molecules in the ER that determine retrograde transport versus ER to Golgi transport. Hahn et al. (35) also reported that calnexin, a membrane-bound and lectin-type chaperone (36), binds to a human meprin ␣ truncated mutant containing a transmembrane domain. This chaperone interaction may be important for correct folding and exit of the glycoprotein from the ER.
The MAM domain in proteins other than meprin does not necessarily function to prevent retrograde transport to the cytosol. For example, when the MAM domain was deleted from receptor protein-tyrosine phosphatase , the protein was well expressed at the plasma membrane (18). The expressed protein lacking the MAM domain, however, lost the ability of specific hemophilic cell-cell interactions in the transfected cells. Thus, while the MAM domain in receptor protein-tyrosine phosphatase plays a role in protein-protein interactions at the cell surface, the phosphatase proteins lacking MAM are not detected as misfolded proteins during biosynthesis. By contrast, meprin mutants lacking MAM are detected as misfolded proteins by the "quality control" systems of the ER. This implies a dependence of the protease domain on the MAM domain to fold correctly.
These results indicate a similarity of the astacin family enzymes to the subtilisin family of proteases (37). Mammalian proprotein convertases are multidomain subtilisin-like endopeptidases with noncatalytic domains COOH-terminal to the catalytic domain. Furin, for example, is composed of signal and prosequences NH 2 -terminal to the serine-type protease domain, with a COOH-terminal homo B domain of about 140 amino acids and a cysteine-rich domain. The homo B domain, which is absent in bacterial subtilisins, is essential for catalytic activity (38). Mutation of the homo B domain results in loss of catalytic activity and miss-sorting of the convertase in the secretory pathway. Astacin family and subtilisin family members have evolved so that there is an important interdependence of domains for correct folding. The complex interactions between domains and putative chaperones add another dimension to the regulation of these proteinases.